Sign in

Blog · Board and management reporting

What makes a change an insight: the three tests behind a weekly sales digest

How to decide which changes in sales data are worth telling a manager about on Monday: the materiality test, the explanation test and the value test, applied to coverage, revenue, share and range week on week, with the digest format that survives the meeting.

The short answerA change is an insight when it passes three tests: it is material against a threshold the team set, it is explained by the rows that caused it, and it has a value attached. Coverage down six points because two reps were out and their routes were not reassigned, worth $48k of skipped stores, is an insight. Coverage down one point is noise. Apply the tests to every measure every week and the digest writes itself.

Every analytics tool can list what changed. Almost none can say which changes matter. A weekly digest that lists forty movements is a report; one that lists three with the reason and the value is an insight. This guide sets out the three tests that separate them, and the digest format that comes out the other side.

The three tests

Materiality. The change is larger than a threshold set for that measure. Coverage moved six points, not one. Revenue in a region moved ten percent, not two.

Explanation. The change is explained by identifiable rows: the accounts, reps, territories or products behind it. Coverage fell because two reps were out and 41 stores were not visited. Not "coverage fell".

Value. The change has a figure attached that says what it is worth: the revenue of the skipped stores, the gap at the region's rate, the ARR of the accounts that went quiet.

A change that fails any test is not in the digest.

Applying them

Every week, for every measure and every level of the roll-up:

  1. Compute the measure for this week and the prior week, and against plan where there is one.
  2. Compute the change and compare it to the measure's threshold.
  3. For changes over the threshold, decompose into the rows that caused it: which members of the next level down moved most.
  4. Value the change at the rate the measure carries.
  5. Rank the surviving changes by value. Keep the top three to five.

A worked digest

Monday, one sales manager, one region.

Signal Measure Change Explanation Value
▼ Midwest coverage −6 pts week on week Two reps out, routes 12 and 15 not reassigned, 41 stores skipped $48k of usual weekly sales
▲ Range gap, South Stores under half range rose to 41 9 new stores on route 7 stocking cola only $31k a month at region average
● Definition change Net revenue moved to v4 Finance excluded a credit category from 1 July All figures below use v4

Three lines. Each has a reason and a number. The manager's Monday is reassigning two routes and calling one rep, and both were decided before the meeting.

What the third line is doing

Not every insight is a sales change. A definition change is material to how the reader interprets every other line, it is explained, and its value is that the numbers are comparable. It belongs in the digest so that nobody asks in the meeting why revenue looks lower than last month's deck.

Where it goes wrong

Thresholds set once. A threshold from a quiet quarter fires every week in a busy one. Recalibrate from history on a schedule.

Explanation stops at the level. "The Midwest is down" is not an explanation; "routes 12 and 15" is. Decompose to the level where somebody can act.

Value omitted. Without a figure, the reader ranks the insights by how they sound. With one, they rank themselves.

Everything to everyone. A rep who receives the region's digest ignores it. Scope the digest to the reader's own rows.

Every Monday, three lines

Mapped once, the weekly file produces the measures, the tests run, and the digest for each reader is three to five lines with the rows behind each. Covirage does this on the weekly export as it is. The weekly insights page describes it.

Questions people ask

How many insights should a weekly digest contain?

Three to five. More than that and nobody reads past the third. The tests are what keep it short: a change that fails any of them is not in the digest, however interesting it looks.

How is materiality set?

Per measure, by the team, from history. A threshold that fires every week is too low; one that never fires is too high. Start at two standard deviations of the measure's week-on-week change and adjust from what the team actually acts on.

Who does the digest go to?

Each person gets the insights for their own scope: a rep gets their accounts, a manager gets the team. The same three tests, applied to a narrower set of rows.